×

Logic of minimal separation in causal networks. (English. Russian original) Zbl 1307.94137

Cybern. Syst. Anal. 49, No. 2, 191-200 (2013); translation from Kibern. Sist. Anal. 2013, No. 2, 36-47 (2013).
Summary: New logical properties and implications are revealed on a subset of pairwise Markov properties satisfied in causal networks. The results obtained characterize a wide class of graphical models including mixed graphs and cyclic digraphs. The following three kinds of separators are defined: minimal, locally minimal, and non-redundant. This article also formulates necessary requirements on members of a non-redundant separator and principles of forming non-redundant separators from elementary (in)dependency facts.

MSC:

94C15 Applications of graph theory to circuits and networks
62A09 Graphical methods in statistics

Software:

TETRAD
PDFBibTeX XMLCite
Full Text: DOI

References:

[1] J. Pearl, Causality: Models, Reasoning, and Inference, Cambridge Univ. Press, Cambridge (2000). · Zbl 0959.68116
[2] P. Spirtes, C. Glymour, and R. Scheines, Causation, Prediction, and Search, 2nd Ed., MIT Press, New York (2001). · Zbl 0981.62001
[3] R. Scheines, P. Spirtes, C. Glymour et al., ”The TETRAD project: Constraint based aids to causal model specification,” Multivar. Behavior. Res., 33, No 1, 65–118 (1998).
[4] T. Richardson and P. Spirtes, ”Ancestral graph Markov models,” Ann. Statistics, 30, No. 4, 962–1030 (2002). · Zbl 1033.60008
[5] P. Spirtes, C. Meek, and T. Richardson, ”An algorithm for causal inference in the presence of latent variables and selection bias,” in: Computation, Causation, and Discovery, AAAI Press, Menlo Park (CA) (1999), pp. 211–252.
[6] P. Spirtes, ”Directed cyclic graphical representations of feedback models,” in: Proc. 11th Conf. on Uncertainty in Artificial Intelligence, Morgan Kaufmann, San Francisco (CA) (1995), pp. 491–498.
[7] A. S. Balabanov, ”Minimal separators in dependency structures: Properties and identification,” Cybernetics and Systems Analysis, 44, No. 6, 803–815 (2008). · Zbl 1179.68104
[8] A. S. Balabanov, ”Construction of d-separators in a dependency system,” Cybernetics and Systems Analysis, 45, No. 5, 703–713 (2009). · Zbl 1216.05045
[9] A. S. Balabanov, A. S. Gapyeyev, A. M. Gupal, and S. S. Rzhepetskiy, ”Fast algorithm for learning the Bayesian networks from data,” Probl. Upravl. Inf., No. 5, 73–80 (2011).
[10] O. S. Balabanov, ”Acceleration of algorithms for Bayesian network recovery: Adaptation to loopfree structures,” Probl. Progr., No. 1, 63–69 (2011).
[11] O. S. Balabanov, ”Acceleration of inductive inference of causal diagrams,” in: Proc. 4th Intern. Workshop on Inductive Modeling (ICIM’2011), Kyiv (2011), pp. 16–21.
[12] T. Verma and J. Pearl, ”Causal networks: Semantics and expressiveness,” in: Proc. 4th Conf. on Uncertainty in Artificial Intelligence (UAI-88), Elsevier, New York (1988), pp. 352–359.
[13] R. M. Neal, ”On deducing conditional independence from d-separation in causal graphs with feedback,” J. Artif. Intellig. Res., 12, 87–91 (2000). · Zbl 0943.68123
[14] O. S. Balabanov, ”Rules of selection of separators in Bayesian networks,” Probl. Progr., No. 4, 22–33 (2007).
[15] A. S. Balabanov, ”Structure recovery of systems of probabilistic dependencies from data: Apparatus of genotypes of variables,” Probl. Upravl. Inf., No 2, 91–99 (2003).
[16] J. Tian, A. Paz, and J. Pearl, ”Finding minimal d-separators,” Tech. Rep., R-254, Computer Sci. Dep., Univ. of California, Los Angeles (CA) (1998).
[17] A. S. Balabanov, ”On the logic of independence of causal diagrams,” in: Computer Mathematics, No. 2, V. M. Glushkov Institute of Cybernetics, Kyiv (2009), pp. 109–115.
[18] S. L. Lauritzen and T. S. Richardson, ”Chain graph models and their causal interpretations,” J. Royal Statist. Soc., B-64, 321–361 (2002). · Zbl 1090.62103
This reference list is based on information provided by the publisher or from digital mathematics libraries. Its items are heuristically matched to zbMATH identifiers and may contain data conversion errors. In some cases that data have been complemented/enhanced by data from zbMATH Open. This attempts to reflect the references listed in the original paper as accurately as possible without claiming completeness or a perfect matching.